Well, suppose one of the operators wanted to cheat the system(they are the only ones who could). There are only two ways he could do it, collecting money that hasn't really been sent or not dispersing money that has been sent.
Say Operator A fraudulently 'sends' $100 to be picked up at Operator B's place. Then A's compatriot goes to B and picks up the dough. Has Operator A made any money? No because They generally use a running tally of debt and repayment to settle up, rather than any money actually changing hands. In other words B is out a hundred bucks but he has also debited A for that amount, its a wash. If A's compatriot makes off with the money then it is A that gets shafted.
As for the other way, not dispersing money that has been sent, if I send money and it never gets to the intended recepient, Cousin Akbar will call me up and tell me so and I will go all over town telling people not to send money via Mid-Eastern Union. Only the honest money senders would survive in an industry so totally predicated on trust.
I guess there is a third way, thats quitting the business while I'm in the red on the running tally and not settling my debt before I run off to Fiji(and even then its only by that amount that you've benefited and even this risk you could minimize by never letting the tallies get too imbalanced). I suppose staffing the network with blood relatives would also cut down on the flight risk.
The article says that the faraway recipient, is identified by name or just a code number. A secret code could make fraudulent receiving of money impossible. Say for instance when Operator A sends Operator B the message to disperse X amount of money to me he sends it encrypted in my PGP public key block, only when I(the intended receiver) come along with my private key can the message be decoded and Operator B knows without a shadow of a doubt that I am the legitimate receiver.
When I read the article earlier, I was under the impression that there was an any-to-any exchange: any broker could communicate with any other and "send money" directly; I didn't (and still don't) see how such a system could work with very many nodes. Every agent would need to know how much every other agent owed him or vice versa, and every agent would have to be able to authenticate and trust every other. Neither of these concepts seems likely.
On the other hand, if the system were set up like FIDOnet, it could probably be made to work with each broker having to keep track of only a small number of other brokers. The brokers would have to know how to route funds through distant brokers, but would only have to vouch for the honesty of those they dealt with directly.
If the system were a many-to-many arrangement, it would probably be possible (and very useful) for the U.S. to infiltrate it and/or use SIGINT to create bogus transactions. Under the FIDOnet-style arrangement, however, this would be very much more difficult. Also, in a many-to-many system, capture and analysis of brokers' records would allow a certain amount of analysis of money flow. Even if the money flowing into and out of most nodes balanced out, most node-to-node links would not be so well balanced. Under a FIDOnet-style system, however, most of the links are going to balance incoming and outgoing money flows pretty well and analysis of money flow will yield very little information.
Gotta admit--sounds like a pretty clever system.